454 PROFESSOR W. DITTMAR ON THE BEHAVIOUR OF THE 



analysis (as quoted under V.) had, on this account, given too high a result for Li, his 

 determinations of the ratio LiCl : LiN0 3 would have done the reverse, while in reality 

 the two results are almost identical. It would be ludicrous for me to try and minimise 

 the strength of Stas's evidence against the correctness of my number, but, then, I cannot 

 get over the fact, that my substance "Q," although proved to contain free Li 2 0, and, in 

 all probability not absolutely free of foreign metals, in three well-agreeing analyses, gave 

 percentages of carbonic acid corresponding to Li = 6*981, 7'004, and 6*997 respectively, 

 without allowing for the free lithia, or for impurities. If my preparations had been 

 absolutely free of magnesia, soda, &c, my value for Li would have become less than it 

 did. If I have gone wrong, it can have been only through two causes : — My carbonate, 

 having been produced in carbonic-acid atmospheres of from 30 to 60 inches tension, may 

 have contained surplus C0 2 , perhaps in the form of a pyro-carbonate Li 2 C0 3 + C0 2 ; or 

 my method for the determination of the carbonic acid may be infected with a constant 

 positive error. To settle these doubts, I caused Mr Anderson to make a few experiments 

 with perfectly pure carbonate of soda according to exactly the same method as had been 

 employed in those lithia experiments which served for the calculation of the atomic weight. 



We had already ascertained that carbonate of soda can be fused in an atmosphere of 

 carbonic acid without appreciable loss of carbonic acid (see next section). 



The results of these special experiments on carbonate of soda are given in the 

 following table: — 





I. 



II. 



III. 



IV. 



C0 2 -pressure, 



60-0 



59-7 



29-2 



29-2 inch. 



Substance analysed, 



8-7906 



5-5292 



7-2228 



4-9794 grru 



C0 2 obtained, 



3-6446 



2-2922 



2-9960 



2-0669 



Percentage of C0 2 , 



41-460 



41-456 



41-480 



41-509 



Calculated value of Na 2 0, 



62-126 



62136 



62-076 



62-001 



Hence, mean value of Na^O = 62*0847, and Na= 23*042, which is a very fair approxi- 

 mation to Stas's value 23*053. 



With these facts before me, I feel inclined to look upon Li= 6*89 as being at present 

 the most probable value of this constant. 



Carbonate of Soda. 



The earlier experiments on this carbonate were made with a supply of Natrum 

 carbonicum purissimum from Trommsdorff of Erfurt, which, as a qualitative analysis 

 showed, really ivas very pure. But it subsequently appeared to me that it would be 

 more satisfactory to work with an absolutely pure salt ; I accordingly, for the 

 subsequent experiments, prepared a still purer salt in the following manner: — 



A quantity of purest soda-crystals ("purissimum" from Erfurt) was dissolved in 

 water in a large nickel vessel, and almost, but not quite, neutralised with pure oxalic 

 acid in the heat, to produce neutral oxalate, which separated out on cooling as a crystal- 

 line powder. As neutral oxalate of potash is easily soluble in water, the trace of 



